By mimicking a follow or creating many different aliases with destructive activities, harmful social bots make bogus messaging and automates their social relations. Additionally, aggressive chatbots send requests from interactive web members to dangerous combination by using shorter hazardous URLs in retweets. As a result, many of the most important tasks in the Transmission of tweeting is to distinguish between malicious fake accounts and legitimate individuals. It requires less effort to find problematic social chatbots using Website address statistics (such as URL redirecting, prevalence of cloned URLs, and spam elements in URLs) than it does using social diagram attributes (which rely on the social interactions of users). Additionally, malicious chatbots are just unable to try to alter URL redirects routes. In this paper, we provide an ensembles of ml algorithms using Web address characteristics for detecting trustworthy social web members (users). Our proposed method provides an ensemble of deep learning systems for categorization and forecasting of online communities to guard against hostile social bot assaults. Experiments were conducted using social web data sets, and the suggested individual's efficiency was tested in terms of specificity, recall, and F-score. To enhance the accuracy, the suggested technique additionally uses a k-fold validation mechanism.